@knurd42 IIUC: Just fetching/cloning anything shallow would be problematic server-load wise then. And then also updating existing clones would (all individual branch HEADs, customized for each state)... ?
Good to be aware about it, but yeah, I think for hammering with CI this doesn't make sense anyway - should be cached CI-side or using plain archives of releases.
Also consider that server-side caching of pack objects which should be hit if it's the same tag for the shallow clone.